作业安全管控系统中人员目标识别算法研究

Study on Human Object Identification algorithm in Working Safety Supervision and Control System

为提高无人值守变电站现场作业安全监督管理水平,对作业安全管控系统中人员目标识别算法进行了研究。在传统算法的基础上,从移动目标捕捉和目标定位、目标分割、人员目标识别三个方面进行算法研究,并根据变电站实际环境对算法进行适用性改进。实验数据表明,不考虑复杂背景的情况下,单人员识别算法的识别率很高,但在目标重叠、目标被部分遮挡、有人影、有积水倒影等情况下,识别率有所下降。人员目标识别算法的实用化研究为无人值守变电站的作业安全管控提供了技术基础。

In order to improve the field working safety supervision and management level in unattended substation, the human object identification algorithm in working safety supervision and control system is studied. Based on the traditional algorithm, the algorithm are studied from three respects including the capturing and locating of moving object, the dividing of object and the personnel object identification, and the applicability of the algorithm is modified according to the actual environment of substation. Test data show that the identification rate of single person is very high under the circumstance of not considering complicated background, but the identification rate is decreased under circumstances of the overlap of object, the partial occlusion of object, the shadow of person and the water reflection. The practical study of human object identification algorithm provide technical basis for the working safety supervision and control in unattended substation.